Research On Classification Algorithm Based On Support Vector Machines And Deep Learning

This paper mainly focuses on machine learning algorithm Support Vector Machines (SVMs) and deep learning method Recursive Auto Encoders (RAE). And their develop progress and application is also studied in detail. And we also researched the shortage of these methods and give out our method to promote them. And the SVMs method and the deep learning methods are also combined to enhance the accuracy and efficiency of clustering. The Followings are main contributions:Firstly, a new support vector machine algorithm with training time reduction is proposed to improve the performance of the algorithm. Traditional SVMs method was seriously depending on the nonsense data during building the hyper planes. In our paper we have proposed a novel SVM method based on effective constructing support vectors to solve this problem. The experiment on the famous data shows that our method works well and can save the training time.Secondly, a novel robust SOM clustering method combined with DE-SVM are proposed. The SOM method is firstly applied to extract data feature. Then DE-SVM is used for classification. The experiment on the UCI database proves that our methods can achieve good performance.Thirdly, a new text semantic information classification method based on the combination of SVM and semi-supervised recursive auto-encoder method (RAE) is proposed. The RAE is used to extract the data feature effectively. SVM is adopted for classification. The result shows that our method can have a good performance on different kinds of data set.